[BUG] plugins.leap: safeLabels = [] does not disable auto-jump

[fin444]

Field	Description
Plugin	leap
Nixpkgs	unstable e8057b6 (June 5)
Home Manager	N/A

• [X] I have read the FAQ and my bug is not listed there.

Description

According to the documentation, setting plugins.leap.safeLabels = [] should disable auto-jump, however it does not.

Minimal, Reproducible Example (MRE)

programs.nixvim = {
  enable = true;
  plugins.leap = {
    enable = true;
    safeLabels = [];
  };
};

Working Example of Auto-Jump Disabled

programs.nixvim = {
  enable = true;
  extraPlugins = [ pkgs.vimPlugins.leap-nvim ];
  extraConfigLua = ''
    require("leap").add_default_mappings()
    require("leap").opts.safe_labels = {}
  '';
};

[GaetanLepage]

Indeed, empty lists and attrs are ignored during the translation to lua. If you want to explicitly have an empty lua table, you can do the following:

setLabels.__empty = null;

[MattSturgeon]

If we ever get to a point where all settings options (inc attrs and lists) are default-null, we might be able to change the behaviour to only omit null attrs?

Maybe we could have a configurable toLuaObject in the meantime where we can toggle behaviours on a per-plugin/per-usage basis? :thinking:

helpers.toLuaObject'
  {
    omitNullAttrs = true;
    omitEmptyAttrs = false;
    multiline = false;
  }
  {
    someNullAttr = null;
    someEmptyAttrsAttr = { };
    someEmptyListAttr = [ ];
  }
=>
''
  { ["someEmptyAttrsAttr"] = { }, ["someEmptyListAttr"] = { } }
''

[traxys]

I think there used to be a reason for empty tables, I remember @GaetanLepage had done some stuff around that some time ago. A good test would be removing the empty attrs from a full config, and seeing how it increases.

[GaetanLepage]

I think there used to be a reason for empty tables, I remember @GaetanLepage had done some stuff around that some time ago. A good test would be removing the empty attrs from a full config, and seeing how it increases.

At some point, we did made the switch to stop ignoring empty attrs. It turned out to be bad. We had to rely on submodules (through helpers.mkCompositeOption) everywhere we needed nested options. We rolled-back this change.

[fin444]

Indeed, empty lists and attrs are ignored during the translation to lua. If you want to explicitly have an empty lua table, you can do the following:

setLabels.__empty = null;

Unfortunately safeLabels takes a list, not a table, so this is not an option

[MattSturgeon]

In lua, lists and dicts are both tables.

[MattSturgeon]

You can also do safeLabels.__raw = "{ }" to do it explicitly in raw lua if __empty isn't working for some reason

[fin444]

On the Nix side, safeLabels is a list. So you can't set __raw or __empty

[MattSturgeon]

Can confirm, the type is nullOr (listOf (maybeRaw str)), so neither __raw or __empty can be used.

This is caused by the underlying issue @traxys brought up in an unrelated discussion: defaultNullOpts.mkListOf does not apply maybeRaw to the list itself, it is only applied to list elements (same formkAttrsOf`).

The fact that __empty can't be used with list-type options is also unfortunate. I think the only way we'd get around that is with a custom listOf type, with an extended check function.

[fin444]

Sorry for taking a while to test this. The changes don't seem to work, auto-jumping still occurs when setting safeLabels.__empty = null; or safeLabels.__raw = "{}"; The below is generated into the config, however. Maybe part of how tbl_deep_extend works?

require("leap").opts = vim.tbl_deep_extend("keep", { safe_labels = {} }, require("leap").opts)

[MattSturgeon]

That sounds likely. The upstream examples all use this pattern:

require('leap').opts[opt_name] = opt_value

I think we either need to do that or use a different "deep extend" mode (maybe force?)

[traxys]

I'd rather follow what upstream recommends